EricHuiK
EricHuiK
Each image channel corresponds to a category heat map？like this?
咨询下，一般来说为了批量训练我们需要把训练的图片resize为固定的大小，那么这样会降低字符的表达能力，如果字符太多那么resize后字符会发生变化，有没有可能是不把训练图片resize为固定大小的方式？那么有没有不适用padding的方式呢？
分割单字符识别不一定使用传统方式，对于单字符分割如果可以分割开来能大大提升模型识别准确度和模型所需的训练数据 ------------------ 原始邮件 ------------------ 发件人: "Mr.Fire"
@avBuffer , I'm also going to try tflite IOS deployment
@avBuffer ，deploy tflite model to Android is ok?
java.lang.IllegalArgumentException: Cannot copy between a TensorFlowLite tensor with shape [1, 2535, 4] and a Java object with shape [1, 10647, 4].
you can convert to .pb（savemodel）,only use pytorch==1.6.0,onnx==1.7.0,tensorflow==2.1.0
I think the different results are model different
python scripts/txt2img.py --prompt "a sunset behind a mountain range, vector image" --ddim_eta 1.0 --n_samples 6 --n_iter 10 --H 256 --W 256 --scale 5.0 --ddim_steps 50
RuntimeError: Given groups=1, weight of size [192, 3, 3, 3], expected input[6, 4, 32, 32] to have 3 channels, but got 4 channels instead